Cameron Simpson <[EMAIL PROTECTED]> writes:
[...]
> Pretty much only things explicitly opening /dev/tty.
>
> You may also get it if some clueless piece of code assumes it's talking
> to a tty and does a stty or something like that.
I've written many cluesless pieces of code..But missed this one till
now... hehe
[...]
> | Your guesses were good ones, and the strace worked to tip it off.
> | It turned out to be an elisp message contained in the loaded file,
> | that gets sent to screen, causing the trouble.
>
> When you say "to screen", what precisely does that mean?
The piece of elisp below, when read by emacs will print in the echo
area of a running emacs:
Hello, your fetch command is reading from ~/.gnus.el.
Which will be annihilated when this fetch completes
[wrapped for mail -ed HP]
(message "\nHello, your fetch command is reading from ~/.gnus.el.\n
Which will be annihilated when this fetch completes\n") (sit-for 1)
When read in `batch moce'( echoing to a tty) it prints the same thing
to `screen'. That is, to an xterm in which the batch command was run.
> | So I guess the cure is to make sure any things that will write to
> | screen are redirected. Doesn't seem to include echo statements
> | though.
What I meant there was, I don't see much difference between the elisp
code writing to xterm screen, and `echo "Something"' doing the same
thing. But apparently Cron sees a difference there.
> Shouldn't be necessary - the output gets logged (for you email) anyway;
> redirecting elswhere shouldn't really help (except that /dev/null is
> a character device, and so may be enough like a tty to make the elisp
> happy).
>
> | Do you have a rough idea what things will turn up in this kind of error?
>
> The strace should show an explicit example.
It showed the example above.
_______________________________________________
Redhat-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list
